{"id":"clm_E5D93F13759D","entity":"openai","predicate":"model_release","value":"InstructGPT","value_json":{"feed":"openai-news","item":"https://openai.com/index/instruction-following","tags":{"v":2,"ai":true,"by":"claude-haiku-5-5","on":"2026-11-03","themes":["safety_policy","reasoning"],"model_tier":"frontier","release_kind":"model"},"check":{"on":"2026-11-04","url":"https://openai.com/index/instruction-following","read":"feed","pages":1,"quote":"These InstructGPT models, which are trained with humans in the loop, are now deployed as the default language models on our API","outcome":"checked"},"evidence":"We’ve trained language models that are much better at following user intentions than GPT-3 while also making them more truthful and less toxic, using techniques developed through our alignment research. These InstructGPT models, which are trained with humans in the loop, are now deployed as the defa","extracted_by":"claude-haiku-5-5","primary_source":true},"valid_from":"2022-01-27","valid_to":null,"published":"2022-01-27","recorded":"2026-11-03","citation":{"plain":"AsOf. \"OpenAI model release: InstructGPT\", as of 2022-01-27. Claim clm_E5D93F13759D, https://asof.is/claim/clm_E5D93F13759D, accessed 2026-11-09. Source: https://openai.com/index/instruction-following.","url":"https://asof.is/claim/clm_E5D93F13759D","accessed":"2026-11-09"},"confidence":0.7,"status":"valid","assertable":true,"superseded_by":null,"supersedes":[],"first_seen_at":"2026-11-03T23:55:38.749992+00:00","updated_at":"2026-11-04T13:19:07.452137+00:00","sources":[{"url":"https://openai.com/index/instruction-following","tier":"primary","publisher":"OpenAI","published_at":"2022-01-27T00:00:00+00:00"}],"evidence":{"by":"page-read","on":"2026-11-04","quote":"These InstructGPT models, which are trained with humans in the loop, are now deployed as the default language models on our API","url":"https://openai.com/index/instruction-following","read":"feed","excerpt":"We’ve trained language models that are much better at following user intentions than GPT-3 while also making them more truthful and less toxic, using techniques developed through our alignment research. These InstructGPT models, which are trained with humans in the loop, are now deployed as the defa"},"legend":{"dates":"valid_from: when the claim took effect (a period figure: the day its period closed). published: the earliest day a source on the record published it (a filing's filing date, a post's date; null when no source carries one). recorded: the day AsOf first held it. A figure can take effect weeks before it is published.","citation":"citation: how to cite the row (plain: one sentence with the value, its as-of date, the claim id and URL, the day answered and the primary source; url: the claim page; accessed: the day answered). APA and BibTeX forms are on the claim page.","why_not":{},"checked_by":{"page-read":"its source page states the value and the date"}}}